CAIN-42 CAIN StudioAgent Autonomy Lease: figures & benchmarks
Evidence for Agent Autonomy Lease: 1. Evidence pack · 2. Answers to buyer questions · 3. Figures & benchmarks · 4. Selling tools · IHVRK kit · Learn it · why these pages exist
Every number on this page is read from the signed transcript inside the IHVRK kit; none is typed in by hand.
5/5
claims matched when recorded
4
attack cases refused or flagged as claimed
58 ms
median round trip (min 46, max 72), incl. network
| # | Kind | Claim | Recorded value | HTTP | ms |
|---|---|---|---|---|---|
| C01 | control | Control: work inside the lease -> Reads, writes and a $30 purchase inside a human-signed level-3 lease: all allowed. | "ALL_ALLOWED" | 200 | 72 |
| C02 | attack | The agent signs itself a level-5 lease with a huge budget -> Self-renewal, an unknown key, and a human lease edited after signing are all refused. | "STOPPED" | 200 | 58 |
| C03 | attack | Many small $45 purchases that add up -> The budget is cumulative: the purchase that crosses $200 is refused (the 'salami' trick). | "STOPPED" | 200 | 46 |
| C04 | attack | A sub-agent asks for more power than its parent -> Delegation can only shrink a lease. A bigger child lease is refused; a smaller one is granted. | "STOPPED" | 200 | 59 |
| C05 | attack | Repeated violations lower the agent's level -> Three refused actions drop the lease from level 3 to 2, so spending is refused afterwards. | "STOPPED" | 200 | 54 |
Copy the table (paste into a spreadsheet)
id kind claim recorded_value expected http ms recorded_at C01 control Control: work inside the lease -> Reads, writes and a $30 purchase inside a human-signed level-3 lease: all allowed. "ALL_ALLOWED" ["ALL_ALLOWED"] 200 72 2026-11-03T22:48:10+00:00 C02 attack The agent signs itself a level-5 lease with a huge budget -> Self-renewal, an unknown key, and a human lease edited after signing are all refused. "STOPPED" ["STOPPED"] 200 58 2026-11-03T22:48:11+00:00 C03 attack Many small $45 purchases that add up -> The budget is cumulative: the purchase that crosses $200 is refused (the 'salami' trick). "STOPPED" ["STOPPED"] 200 46 2026-11-03T22:48:11+00:00 C04 attack A sub-agent asks for more power than its parent -> Delegation can only shrink a lease. A bigger child lease is refused; a smaller one is granted. "STOPPED" ["STOPPED"] 200 59 2026-11-03T22:48:12+00:00 C05 attack Repeated violations lower the agent's level -> Three refused actions drop the lease from level 3 to 2, so spending is refused afterwards. "STOPPED" ["STOPPED"] 200 54 2026-11-03T22:48:12+00:00
Download: figures.tsv · transcript.json (full answers) · whole kit
How to read these numbers
- Recorded 2026-11-03T22:48:13+00:00 against cainstudio.online from our build host; times include the internet round trip.
- They show these inputs, not every possible input. Known misses are on the limits page.
- This is not a load test, an SLA or a third-party benchmark. Re-run it yourself:
python3 verify.py --live.